Model benchmarks
Qwen3.6-35B-A3B-UD-IQ3_S local LLM performance
As of October 2026, Qwen3.6-35B-A3B-UD-IQ3_S runs at up to 11.2 tok/s for local inference (best of 2 community benchmark runs across 1 GPU).
Model size
35B
Peak speed
11.2 tok/s
Average speed
11.2 tok/s
Avg PP
85.6 tok/s
Min memory
13.5 GB
Max context
8,192 tokens
Avg output / run
7,422 tokens
Avg runtime / run
11m 34s
Avg quality
71.4
Benchmark runs
2
GPUs tested
1
Quality by task
Average LLM-judged quality (0–100) with the run-to-run spread shown as a P5–P95 band, overall and for each benchmark task, across all 2 runs. The low and high columns show how much the judge’s score varies between runs, and need at least two runs to display.
| Task | P5 (low) | Avg | P95 (high) |
|---|---|---|---|
| Overall | 68.5 | 71.4 | 74.3 |
| Agent Workflow | 48.3 | 61.1 | 73.9 |
| Code Generation | 60.2 | 65.7 | 71.2 |
| Role Play & Narrative | 79.8 | 83.7 | 87.6 |
| Research & Analysis | 72.5 | 75.2 | 77.9 |
Performance by hardware and tool
Every hardware/tool/quantization combination Qwen3.6-35B-A3B-UD-IQ3_S has been benchmarked on, ranked by peak token generation speed. Last updated October 2026.
| Hardware | Tool | Quant | Best tok/s | Avg tok/s | Memory | Context | Quality | Runs |
|---|---|---|---|---|---|---|---|---|
| AMD Radeon 680M | llama.cpp | — | 11.2 tok/s | 11.2 tok/s | 13.5 GB | 8,192 tokens | 71.4 | 2 |
Benchmark runs
All 2 Qwen3.6-35B-A3B-UD-IQ3_S runs submitted so far — expand one for its hardware, quality breakdown and per-scenario detail.
Frequently asked questions
- Is Qwen3.6-35B-A3B-UD-IQ3_S good for coding?
- In our benchmarks, Qwen3.6-35B-A3B-UD-IQ3_S scores 65.7/100 for coding. It runs at about 11.2 tok/s, so if you want more speed, tensorfold-qwen38-27b is faster (~25.0 tok/s) and still scores well for coding (84.3/100).
- Is Qwen3.6-35B-A3B-UD-IQ3_S good for agentic (tool-using) tasks?
- In our benchmarks, Qwen3.6-35B-A3B-UD-IQ3_S scores 61.1/100 for agentic workflows. It runs at about 11.2 tok/s, so if you want more speed, muse-glimmer:latest is faster (~33.3 tok/s) and still scores well for agentic workflows (91.6/100).
- How fast is Qwen3.6-35B-A3B-UD-IQ3_S for local inference?
- Across 2 community benchmark runs, Qwen3.6-35B-A3B-UD-IQ3_S reaches up to 11.2 tok/s and averages 11.2 tok/s, with the fastest results on AMD Radeon 680M.
- How much memory does Qwen3.6-35B-A3B-UD-IQ3_S need?
- The leanest observed configuration used about 13.5 GB of memory.
- Which tools have been used to run Qwen3.6-35B-A3B-UD-IQ3_S?
- Benchmarks were submitted using llama.cpp. Results are community-contributed and updated as new runs arrive.